drivers/scsi/xen-scsifront.c

   1 /*
   2  * Xen SCSI frontend driver
   3  *
   4  * Copyright (c) 2008, FUJITSU Limited
   5  *
   6  * This program is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU General Public License version 2
   8  * as published by the Free Software Foundation; or, when distributed
   9  * separately from the Linux kernel or incorporated into other
  10  * software packages, subject to the following license:
  11  *
  12  * Permission is hereby granted, free of charge, to any person obtaining a copy
  13  * of this source file (the "Software"), to deal in the Software without
  14  * restriction, including without limitation the rights to use, copy, modify,
  15  * merge, publish, distribute, sublicense, and/or sell copies of the Software,
  16  * and to permit persons to whom the Software is furnished to do so, subject to
  17  * the following conditions:
  18  *
  19  * The above copyright notice and this permission notice shall be included in
  20  * all copies or substantial portions of the Software.
  21  *
  22  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  23  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  24  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  25  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  26  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  27  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  28  * IN THE SOFTWARE.
  29  */
  30
  31 #include <linux/module.h>
  32 #include <linux/kernel.h>
  33 #include <linux/device.h>
  34 #include <linux/wait.h>
  35 #include <linux/interrupt.h>
  36 #include <linux/mutex.h>
  37 #include <linux/spinlock.h>
  38 #include <linux/sched.h>
  39 #include <linux/blkdev.h>
  40 #include <linux/pfn.h>
  41 #include <linux/slab.h>
  42 #include <linux/bitops.h>
  43
  44 #include <scsi/scsi_cmnd.h>
  45 #include <scsi/scsi_device.h>
  46 #include <scsi/scsi.h>
  47 #include <scsi/scsi_host.h>
  48
  49 #include <xen/xen.h>
  50 #include <xen/xenbus.h>
  51 #include <xen/grant_table.h>
  52 #include <xen/events.h>
  53 #include <xen/page.h>
  54
  55 #include <xen/interface/grant_table.h>
  56 #include <xen/interface/io/vscsiif.h>
  57 #include <xen/interface/io/protocols.h>
  58
  59 #include <asm/xen/hypervisor.h>
  60
  61
  62 #define GRANT_INVALID_REF       0
  63
  64 #define VSCSIFRONT_OP_ADD_LUN   1
  65 #define VSCSIFRONT_OP_DEL_LUN   2
  66 #define VSCSIFRONT_OP_READD_LUN 3
  67
  68 /* Tuning point. */
  69 #define VSCSIIF_DEFAULT_CMD_PER_LUN 10
  70 #define VSCSIIF_MAX_TARGET          64
  71 #define VSCSIIF_MAX_LUN             255
  72
  73 #define VSCSIIF_RING_SIZE       __CONST_RING_SIZE(vscsiif, PAGE_SIZE)
  74 #define VSCSIIF_MAX_REQS        VSCSIIF_RING_SIZE
  75
  76 #define vscsiif_grants_sg(_sg)  (PFN_UP((_sg) *         \
  77                                 sizeof(struct scsiif_request_segment)))
  78
  79 struct vscsifrnt_shadow {
  80         /* command between backend and frontend */
  81         unsigned char act;
  82         uint8_t nr_segments;
  83         uint16_t rqid;
  84         uint16_t ref_rqid;
  85
  86         unsigned int nr_grants;         /* number of grants in gref[] */
  87         struct scsiif_request_segment *sg;      /* scatter/gather elements */
  88         struct scsiif_request_segment seg[VSCSIIF_SG_TABLESIZE];
  89
  90         /* Do reset or abort function. */
  91         wait_queue_head_t wq_reset;     /* reset work queue           */
  92         int wait_reset;                 /* reset work queue condition */
  93         int32_t rslt_reset;             /* reset response status:     */
  94                                         /* SUCCESS or FAILED or:      */
  95 #define RSLT_RESET_WAITING      0
  96 #define RSLT_RESET_ERR          -1
  97
  98         /* Requested struct scsi_cmnd is stored from kernel. */
  99         struct scsi_cmnd *sc;
 100         int gref[vscsiif_grants_sg(SG_ALL) + SG_ALL];
 101 };
 102
 103 struct vscsifrnt_info {
 104         struct xenbus_device *dev;
 105
 106         struct Scsi_Host *host;
 107         int host_active;
 108
 109         unsigned int evtchn;
 110         unsigned int irq;
 111
 112         grant_ref_t ring_ref;
 113         struct vscsiif_front_ring ring;
 114         struct vscsiif_response ring_rsp;
 115
 116         spinlock_t shadow_lock;
 117         DECLARE_BITMAP(shadow_free_bitmap, VSCSIIF_MAX_REQS);
 118         struct vscsifrnt_shadow *shadow[VSCSIIF_MAX_REQS];
 119
 120         /* Following items are protected by the host lock. */
 121         wait_queue_head_t wq_sync;
 122         wait_queue_head_t wq_pause;
 123         unsigned int wait_ring_available:1;
 124         unsigned int waiting_pause:1;
 125         unsigned int pause:1;
 126         unsigned callers;
 127
 128         char dev_state_path[64];
 129         struct task_struct *curr;
 130 };
 131
 132 static DEFINE_MUTEX(scsifront_mutex);
 133
 134 static void scsifront_wake_up(struct vscsifrnt_info *info)
 135 {
 136         info->wait_ring_available = 0;
 137         wake_up(&info->wq_sync);
 138 }
 139
 140 static int scsifront_get_rqid(struct vscsifrnt_info *info)
 141 {
 142         unsigned long flags;
 143         int free;
 144
 145         spin_lock_irqsave(&info->shadow_lock, flags);
 146
 147         free = find_first_bit(info->shadow_free_bitmap, VSCSIIF_MAX_REQS);
 148         __clear_bit(free, info->shadow_free_bitmap);
 149
 150         spin_unlock_irqrestore(&info->shadow_lock, flags);
 151
 152         return free;
 153 }
 154
 155 static int _scsifront_put_rqid(struct vscsifrnt_info *info, uint32_t id)
 156 {
 157         int empty = bitmap_empty(info->shadow_free_bitmap, VSCSIIF_MAX_REQS);
 158
 159         __set_bit(id, info->shadow_free_bitmap);
 160         info->shadow[id] = NULL;
 161
 162         return empty || info->wait_ring_available;
 163 }
 164
 165 static void scsifront_put_rqid(struct vscsifrnt_info *info, uint32_t id)
 166 {
 167         unsigned long flags;
 168         int kick;
 169
 170         spin_lock_irqsave(&info->shadow_lock, flags);
 171         kick = _scsifront_put_rqid(info, id);
 172         spin_unlock_irqrestore(&info->shadow_lock, flags);
 173
 174         if (kick)
 175                 scsifront_wake_up(info);
 176 }
 177
 178 static int scsifront_do_request(struct vscsifrnt_info *info,
 179                                 struct vscsifrnt_shadow *shadow)
 180 {
 181         struct vscsiif_front_ring *ring = &(info->ring);
 182         struct vscsiif_request *ring_req;
 183         struct scsi_cmnd *sc = shadow->sc;
 184         uint32_t id;
 185         int i, notify;
 186
 187         if (RING_FULL(&info->ring))
 188                 return -EBUSY;
 189
 190         id = scsifront_get_rqid(info);  /* use id in response */
 191         if (id >= VSCSIIF_MAX_REQS)
 192                 return -EBUSY;
 193
 194         info->shadow[id] = shadow;
 195         shadow->rqid = id;
 196
 197         ring_req = RING_GET_REQUEST(&(info->ring), ring->req_prod_pvt);
 198         ring->req_prod_pvt++;
 199
 200         ring_req->rqid        = id;
 201         ring_req->act         = shadow->act;
 202         ring_req->ref_rqid    = shadow->ref_rqid;
 203         ring_req->nr_segments = shadow->nr_segments;
 204
 205         ring_req->id      = sc->device->id;
 206         ring_req->lun     = sc->device->lun;
 207         ring_req->channel = sc->device->channel;
 208         ring_req->cmd_len = sc->cmd_len;
 209
 210         BUG_ON(sc->cmd_len > VSCSIIF_MAX_COMMAND_SIZE);
 211
 212         memcpy(ring_req->cmnd, sc->cmnd, sc->cmd_len);
 213
 214         ring_req->sc_data_direction   = (uint8_t)sc->sc_data_direction;
 215         ring_req->timeout_per_command = scsi_cmd_to_rq(sc)->timeout / HZ;
 216
 217         for (i = 0; i < (shadow->nr_segments & ~VSCSIIF_SG_GRANT); i++)
 218                 ring_req->seg[i] = shadow->seg[i];
 219
 220         RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(ring, notify);
 221         if (notify)
 222                 notify_remote_via_irq(info->irq);
 223
 224         return 0;
 225 }
 226
 227 static void scsifront_gnttab_done(struct vscsifrnt_info *info,
 228                                   struct vscsifrnt_shadow *shadow)
 229 {
 230         int i;
 231
 232         if (shadow->sc->sc_data_direction == DMA_NONE)
 233                 return;
 234
 235         for (i = 0; i < shadow->nr_grants; i++) {
 236                 if (unlikely(gnttab_query_foreign_access(shadow->gref[i]))) {
 237                         shost_printk(KERN_ALERT, info->host, KBUILD_MODNAME
 238                                      "grant still in use by backend\n");
 239                         BUG();
 240                 }
 241                 gnttab_end_foreign_access(shadow->gref[i], 0, 0UL);
 242         }
 243
 244         kfree(shadow->sg);
 245 }
 246
 247 static void scsifront_cdb_cmd_done(struct vscsifrnt_info *info,
 248                                    struct vscsiif_response *ring_rsp)
 249 {
 250         struct vscsifrnt_shadow *shadow;
 251         struct scsi_cmnd *sc;
 252         uint32_t id;
 253         uint8_t sense_len;
 254         int result;
 255
 256         id = ring_rsp->rqid;
 257         shadow = info->shadow[id];
 258         sc = shadow->sc;
 259
 260         BUG_ON(sc == NULL);
 261
 262         scsifront_gnttab_done(info, shadow);
 263         scsifront_put_rqid(info, id);
 264
 265         result = ring_rsp->rslt;
 266         if (result >> 24)
 267                 set_host_byte(sc, DID_ERROR);
 268         else
 269                 set_host_byte(sc, host_byte(result));
 270         set_status_byte(sc, result & 0xff);
 271         scsi_set_resid(sc, ring_rsp->residual_len);
 272
 273         sense_len = min_t(uint8_t, VSCSIIF_SENSE_BUFFERSIZE,
 274                           ring_rsp->sense_len);
 275
 276         if (sense_len)
 277                 memcpy(sc->sense_buffer, ring_rsp->sense_buffer, sense_len);
 278
 279         scsi_done(sc);
 280 }
 281
 282 static void scsifront_sync_cmd_done(struct vscsifrnt_info *info,
 283                                     struct vscsiif_response *ring_rsp)
 284 {
 285         uint16_t id = ring_rsp->rqid;
 286         unsigned long flags;
 287         struct vscsifrnt_shadow *shadow = info->shadow[id];
 288         int kick;
 289
 290         spin_lock_irqsave(&info->shadow_lock, flags);
 291         shadow->wait_reset = 1;
 292         switch (shadow->rslt_reset) {
 293         case RSLT_RESET_WAITING:
 294                 shadow->rslt_reset = ring_rsp->rslt;
 295                 break;
 296         case RSLT_RESET_ERR:
 297                 kick = _scsifront_put_rqid(info, id);
 298                 spin_unlock_irqrestore(&info->shadow_lock, flags);
 299                 kfree(shadow);
 300                 if (kick)
 301                         scsifront_wake_up(info);
 302                 return;
 303         default:
 304                 shost_printk(KERN_ERR, info->host, KBUILD_MODNAME
 305                              "bad reset state %d, possibly leaking %u\n",
 306                              shadow->rslt_reset, id);
 307                 break;
 308         }
 309         spin_unlock_irqrestore(&info->shadow_lock, flags);
 310
 311         wake_up(&shadow->wq_reset);
 312 }
 313
 314 static void scsifront_do_response(struct vscsifrnt_info *info,
 315                                   struct vscsiif_response *ring_rsp)
 316 {
 317         if (WARN(ring_rsp->rqid >= VSCSIIF_MAX_REQS ||
 318                  test_bit(ring_rsp->rqid, info->shadow_free_bitmap),
 319                  "illegal rqid %u returned by backend!\n", ring_rsp->rqid))
 320                 return;
 321
 322         if (info->shadow[ring_rsp->rqid]->act == VSCSIIF_ACT_SCSI_CDB)
 323                 scsifront_cdb_cmd_done(info, ring_rsp);
 324         else
 325                 scsifront_sync_cmd_done(info, ring_rsp);
 326 }
 327
 328 static int scsifront_ring_drain(struct vscsifrnt_info *info)
 329 {
 330         struct vscsiif_response *ring_rsp;
 331         RING_IDX i, rp;
 332         int more_to_do = 0;
 333
 334         rp = info->ring.sring->rsp_prod;
 335         rmb();  /* ordering required respective to dom0 */
 336         for (i = info->ring.rsp_cons; i != rp; i++) {
 337                 ring_rsp = RING_GET_RESPONSE(&info->ring, i);
 338                 scsifront_do_response(info, ring_rsp);
 339         }
 340
 341         info->ring.rsp_cons = i;
 342
 343         if (i != info->ring.req_prod_pvt)
 344                 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do);
 345         else
 346                 info->ring.sring->rsp_event = i + 1;
 347
 348         return more_to_do;
 349 }
 350
 351 static int scsifront_cmd_done(struct vscsifrnt_info *info)
 352 {
 353         int more_to_do;
 354         unsigned long flags;
 355
 356         spin_lock_irqsave(info->host->host_lock, flags);
 357
 358         more_to_do = scsifront_ring_drain(info);
 359
 360         info->wait_ring_available = 0;
 361
 362         spin_unlock_irqrestore(info->host->host_lock, flags);
 363
 364         wake_up(&info->wq_sync);
 365
 366         return more_to_do;
 367 }
 368
 369 static irqreturn_t scsifront_irq_fn(int irq, void *dev_id)
 370 {
 371         struct vscsifrnt_info *info = dev_id;
 372
 373         while (scsifront_cmd_done(info))
 374                 /* Yield point for this unbounded loop. */
 375                 cond_resched();
 376
 377         return IRQ_HANDLED;
 378 }
 379
 380 static void scsifront_finish_all(struct vscsifrnt_info *info)
 381 {
 382         unsigned i;
 383         struct vscsiif_response resp;
 384
 385         scsifront_ring_drain(info);
 386
 387         for (i = 0; i < VSCSIIF_MAX_REQS; i++) {
 388                 if (test_bit(i, info->shadow_free_bitmap))
 389                         continue;
 390                 resp.rqid = i;
 391                 resp.sense_len = 0;
 392                 resp.rslt = DID_RESET << 16;
 393                 resp.residual_len = 0;
 394                 scsifront_do_response(info, &resp);
 395         }
 396 }
 397
 398 static int map_data_for_request(struct vscsifrnt_info *info,
 399                                 struct scsi_cmnd *sc,
 400                                 struct vscsifrnt_shadow *shadow)
 401 {
 402         grant_ref_t gref_head;
 403         struct page *page;
 404         int err, ref, ref_cnt = 0;
 405         int grant_ro = (sc->sc_data_direction == DMA_TO_DEVICE);
 406         unsigned int i, off, len, bytes;
 407         unsigned int data_len = scsi_bufflen(sc);
 408         unsigned int data_grants = 0, seg_grants = 0;
 409         struct scatterlist *sg;
 410         struct scsiif_request_segment *seg;
 411
 412         if (sc->sc_data_direction == DMA_NONE || !data_len)
 413                 return 0;
 414
 415         scsi_for_each_sg(sc, sg, scsi_sg_count(sc), i)
 416                 data_grants += PFN_UP(sg->offset + sg->length);
 417
 418         if (data_grants > VSCSIIF_SG_TABLESIZE) {
 419                 if (data_grants > info->host->sg_tablesize) {
 420                         shost_printk(KERN_ERR, info->host, KBUILD_MODNAME
 421                              "Unable to map request_buffer for command!\n");
 422                         return -E2BIG;
 423                 }
 424                 seg_grants = vscsiif_grants_sg(data_grants);
 425                 shadow->sg = kcalloc(data_grants,
 426                         sizeof(struct scsiif_request_segment), GFP_ATOMIC);
 427                 if (!shadow->sg)
 428                         return -ENOMEM;
 429         }
 430         seg = shadow->sg ? : shadow->seg;
 431
 432         err = gnttab_alloc_grant_references(seg_grants + data_grants,
 433                                             &gref_head);
 434         if (err) {
 435                 kfree(shadow->sg);
 436                 shost_printk(KERN_ERR, info->host, KBUILD_MODNAME
 437                              "gnttab_alloc_grant_references() error\n");
 438                 return -ENOMEM;
 439         }
 440
 441         if (seg_grants) {
 442                 page = virt_to_page(seg);
 443                 off = offset_in_page(seg);
 444                 len = sizeof(struct scsiif_request_segment) * data_grants;
 445                 while (len > 0) {
 446                         bytes = min_t(unsigned int, len, PAGE_SIZE - off);
 447
 448                         ref = gnttab_claim_grant_reference(&gref_head);
 449                         BUG_ON(ref == -ENOSPC);
 450
 451                         gnttab_grant_foreign_access_ref(ref,
 452                                 info->dev->otherend_id,
 453                                 xen_page_to_gfn(page), 1);
 454                         shadow->gref[ref_cnt] = ref;
 455                         shadow->seg[ref_cnt].gref   = ref;
 456                         shadow->seg[ref_cnt].offset = (uint16_t)off;
 457                         shadow->seg[ref_cnt].length = (uint16_t)bytes;
 458
 459                         page++;
 460                         len -= bytes;
 461                         off = 0;
 462                         ref_cnt++;
 463                 }
 464                 BUG_ON(seg_grants < ref_cnt);
 465                 seg_grants = ref_cnt;
 466         }
 467
 468         scsi_for_each_sg(sc, sg, scsi_sg_count(sc), i) {
 469                 page = sg_page(sg);
 470                 off = sg->offset;
 471                 len = sg->length;
 472
 473                 while (len > 0 && data_len > 0) {
 474                         /*
 475                          * sg sends a scatterlist that is larger than
 476                          * the data_len it wants transferred for certain
 477                          * IO sizes.
 478                          */
 479                         bytes = min_t(unsigned int, len, PAGE_SIZE - off);
 480                         bytes = min(bytes, data_len);
 481
 482                         ref = gnttab_claim_grant_reference(&gref_head);
 483                         BUG_ON(ref == -ENOSPC);
 484
 485                         gnttab_grant_foreign_access_ref(ref,
 486                                 info->dev->otherend_id,
 487                                 xen_page_to_gfn(page),
 488                                 grant_ro);
 489
 490                         shadow->gref[ref_cnt] = ref;
 491                         seg->gref   = ref;
 492                         seg->offset = (uint16_t)off;
 493                         seg->length = (uint16_t)bytes;
 494
 495                         page++;
 496                         seg++;
 497                         len -= bytes;
 498                         data_len -= bytes;
 499                         off = 0;
 500                         ref_cnt++;
 501                 }
 502         }
 503
 504         if (seg_grants)
 505                 shadow->nr_segments = VSCSIIF_SG_GRANT | seg_grants;
 506         else
 507                 shadow->nr_segments = (uint8_t)ref_cnt;
 508         shadow->nr_grants = ref_cnt;
 509
 510         return 0;
 511 }
 512
 513 static int scsifront_enter(struct vscsifrnt_info *info)
 514 {
 515         if (info->pause)
 516                 return 1;
 517         info->callers++;
 518         return 0;
 519 }
 520
 521 static void scsifront_return(struct vscsifrnt_info *info)
 522 {
 523         info->callers--;
 524         if (info->callers)
 525                 return;
 526
 527         if (!info->waiting_pause)
 528                 return;
 529
 530         info->waiting_pause = 0;
 531         wake_up(&info->wq_pause);
 532 }
 533
 534 static int scsifront_queuecommand(struct Scsi_Host *shost,
 535                                   struct scsi_cmnd *sc)
 536 {
 537         struct vscsifrnt_info *info = shost_priv(shost);
 538         struct vscsifrnt_shadow *shadow = scsi_cmd_priv(sc);
 539         unsigned long flags;
 540         int err;
 541
 542         sc->result = 0;
 543
 544         shadow->sc  = sc;
 545         shadow->act = VSCSIIF_ACT_SCSI_CDB;
 546
 547         spin_lock_irqsave(shost->host_lock, flags);
 548         if (scsifront_enter(info)) {
 549                 spin_unlock_irqrestore(shost->host_lock, flags);
 550                 return SCSI_MLQUEUE_HOST_BUSY;
 551         }
 552
 553         err = map_data_for_request(info, sc, shadow);
 554         if (err < 0) {
 555                 pr_debug("%s: err %d\n", __func__, err);
 556                 scsifront_return(info);
 557                 spin_unlock_irqrestore(shost->host_lock, flags);
 558                 if (err == -ENOMEM)
 559                         return SCSI_MLQUEUE_HOST_BUSY;
 560                 sc->result = DID_ERROR << 16;
 561                 scsi_done(sc);
 562                 return 0;
 563         }
 564
 565         if (scsifront_do_request(info, shadow)) {
 566                 scsifront_gnttab_done(info, shadow);
 567                 goto busy;
 568         }
 569
 570         scsifront_return(info);
 571         spin_unlock_irqrestore(shost->host_lock, flags);
 572
 573         return 0;
 574
 575 busy:
 576         scsifront_return(info);
 577         spin_unlock_irqrestore(shost->host_lock, flags);
 578         pr_debug("%s: busy\n", __func__);
 579         return SCSI_MLQUEUE_HOST_BUSY;
 580 }
 581
 582 /*
 583  * Any exception handling (reset or abort) must be forwarded to the backend.
 584  * We have to wait until an answer is returned. This answer contains the
 585  * result to be returned to the requestor.
 586  */
 587 static int scsifront_action_handler(struct scsi_cmnd *sc, uint8_t act)
 588 {
 589         struct Scsi_Host *host = sc->device->host;
 590         struct vscsifrnt_info *info = shost_priv(host);
 591         struct vscsifrnt_shadow *shadow, *s = scsi_cmd_priv(sc);
 592         int err = 0;
 593
 594         shadow = kzalloc(sizeof(*shadow), GFP_NOIO);
 595         if (!shadow)
 596                 return FAILED;
 597
 598         shadow->act = act;
 599         shadow->rslt_reset = RSLT_RESET_WAITING;
 600         shadow->sc = sc;
 601         shadow->ref_rqid = s->rqid;
 602         init_waitqueue_head(&shadow->wq_reset);
 603
 604         spin_lock_irq(host->host_lock);
 605
 606         for (;;) {
 607                 if (scsifront_enter(info))
 608                         goto fail;
 609
 610                 if (!scsifront_do_request(info, shadow))
 611                         break;
 612
 613                 scsifront_return(info);
 614                 if (err)
 615                         goto fail;
 616                 info->wait_ring_available = 1;
 617                 spin_unlock_irq(host->host_lock);
 618                 err = wait_event_interruptible(info->wq_sync,
 619                                                !info->wait_ring_available);
 620                 spin_lock_irq(host->host_lock);
 621         }
 622
 623         spin_unlock_irq(host->host_lock);
 624         err = wait_event_interruptible(shadow->wq_reset, shadow->wait_reset);
 625         spin_lock_irq(host->host_lock);
 626
 627         if (!err) {
 628                 err = shadow->rslt_reset;
 629                 scsifront_put_rqid(info, shadow->rqid);
 630                 kfree(shadow);
 631         } else {
 632                 spin_lock(&info->shadow_lock);
 633                 shadow->rslt_reset = RSLT_RESET_ERR;
 634                 spin_unlock(&info->shadow_lock);
 635                 err = FAILED;
 636         }
 637
 638         scsifront_return(info);
 639         spin_unlock_irq(host->host_lock);
 640         return err;
 641
 642 fail:
 643         spin_unlock_irq(host->host_lock);
 644         kfree(shadow);
 645         return FAILED;
 646 }
 647
 648 static int scsifront_eh_abort_handler(struct scsi_cmnd *sc)
 649 {
 650         pr_debug("%s\n", __func__);
 651         return scsifront_action_handler(sc, VSCSIIF_ACT_SCSI_ABORT);
 652 }
 653
 654 static int scsifront_dev_reset_handler(struct scsi_cmnd *sc)
 655 {
 656         pr_debug("%s\n", __func__);
 657         return scsifront_action_handler(sc, VSCSIIF_ACT_SCSI_RESET);
 658 }
 659
 660 static int scsifront_sdev_configure(struct scsi_device *sdev)
 661 {
 662         struct vscsifrnt_info *info = shost_priv(sdev->host);
 663         int err;
 664
 665         if (info && current == info->curr) {
 666                 err = xenbus_printf(XBT_NIL, info->dev->nodename,
 667                               info->dev_state_path, "%d", XenbusStateConnected);
 668                 if (err) {
 669                         xenbus_dev_error(info->dev, err,
 670                                 "%s: writing dev_state_path", __func__);
 671                         return err;
 672                 }
 673         }
 674
 675         return 0;
 676 }
 677
 678 static void scsifront_sdev_destroy(struct scsi_device *sdev)
 679 {
 680         struct vscsifrnt_info *info = shost_priv(sdev->host);
 681         int err;
 682
 683         if (info && current == info->curr) {
 684                 err = xenbus_printf(XBT_NIL, info->dev->nodename,
 685                               info->dev_state_path, "%d", XenbusStateClosed);
 686                 if (err)
 687                         xenbus_dev_error(info->dev, err,
 688                                 "%s: writing dev_state_path", __func__);
 689         }
 690 }
 691
 692 static struct scsi_host_template scsifront_sht = {
 693         .module                 = THIS_MODULE,
 694         .name                   = "Xen SCSI frontend driver",
 695         .queuecommand           = scsifront_queuecommand,
 696         .eh_abort_handler       = scsifront_eh_abort_handler,
 697         .eh_device_reset_handler = scsifront_dev_reset_handler,
 698         .slave_configure        = scsifront_sdev_configure,
 699         .slave_destroy          = scsifront_sdev_destroy,
 700         .cmd_per_lun            = VSCSIIF_DEFAULT_CMD_PER_LUN,
 701         .can_queue              = VSCSIIF_MAX_REQS,
 702         .this_id                = -1,
 703         .cmd_size               = sizeof(struct vscsifrnt_shadow),
 704         .sg_tablesize           = VSCSIIF_SG_TABLESIZE,
 705         .proc_name              = "scsifront",
 706 };
 707
 708 static int scsifront_alloc_ring(struct vscsifrnt_info *info)
 709 {
 710         struct xenbus_device *dev = info->dev;
 711         struct vscsiif_sring *sring;
 712         grant_ref_t gref;
 713         int err = -ENOMEM;
 714
 715         /***** Frontend to Backend ring start *****/
 716         sring = (struct vscsiif_sring *)__get_free_page(GFP_KERNEL);
 717         if (!sring) {
 718                 xenbus_dev_fatal(dev, err,
 719                         "fail to allocate shared ring (Front to Back)");
 720                 return err;
 721         }
 722         SHARED_RING_INIT(sring);
 723         FRONT_RING_INIT(&info->ring, sring, PAGE_SIZE);
 724
 725         err = xenbus_grant_ring(dev, sring, 1, &gref);
 726         if (err < 0) {
 727                 free_page((unsigned long)sring);
 728                 xenbus_dev_fatal(dev, err,
 729                         "fail to grant shared ring (Front to Back)");
 730                 return err;
 731         }
 732         info->ring_ref = gref;
 733
 734         err = xenbus_alloc_evtchn(dev, &info->evtchn);
 735         if (err) {
 736                 xenbus_dev_fatal(dev, err, "xenbus_alloc_evtchn");
 737                 goto free_gnttab;
 738         }
 739
 740         err = bind_evtchn_to_irq(info->evtchn);
 741         if (err <= 0) {
 742                 xenbus_dev_fatal(dev, err, "bind_evtchn_to_irq");
 743                 goto free_gnttab;
 744         }
 745
 746         info->irq = err;
 747
 748         err = request_threaded_irq(info->irq, NULL, scsifront_irq_fn,
 749                                    IRQF_ONESHOT, "scsifront", info);
 750         if (err) {
 751                 xenbus_dev_fatal(dev, err, "request_threaded_irq");
 752                 goto free_irq;
 753         }
 754
 755         return 0;
 756
 757 /* free resource */
 758 free_irq:
 759         unbind_from_irqhandler(info->irq, info);
 760 free_gnttab:
 761         gnttab_end_foreign_access(info->ring_ref, 0,
 762                                   (unsigned long)info->ring.sring);
 763
 764         return err;
 765 }
 766
 767 static void scsifront_free_ring(struct vscsifrnt_info *info)
 768 {
 769         unbind_from_irqhandler(info->irq, info);
 770         gnttab_end_foreign_access(info->ring_ref, 0,
 771                                   (unsigned long)info->ring.sring);
 772 }
 773
 774 static int scsifront_init_ring(struct vscsifrnt_info *info)
 775 {
 776         struct xenbus_device *dev = info->dev;
 777         struct xenbus_transaction xbt;
 778         int err;
 779
 780         pr_debug("%s\n", __func__);
 781
 782         err = scsifront_alloc_ring(info);
 783         if (err)
 784                 return err;
 785         pr_debug("%s: %u %u\n", __func__, info->ring_ref, info->evtchn);
 786
 787 again:
 788         err = xenbus_transaction_start(&xbt);
 789         if (err)
 790                 xenbus_dev_fatal(dev, err, "starting transaction");
 791
 792         err = xenbus_printf(xbt, dev->nodename, "ring-ref", "%u",
 793                             info->ring_ref);
 794         if (err) {
 795                 xenbus_dev_fatal(dev, err, "%s", "writing ring-ref");
 796                 goto fail;
 797         }
 798
 799         err = xenbus_printf(xbt, dev->nodename, "event-channel", "%u",
 800                             info->evtchn);
 801
 802         if (err) {
 803                 xenbus_dev_fatal(dev, err, "%s", "writing event-channel");
 804                 goto fail;
 805         }
 806
 807         err = xenbus_transaction_end(xbt, 0);
 808         if (err) {
 809                 if (err == -EAGAIN)
 810                         goto again;
 811                 xenbus_dev_fatal(dev, err, "completing transaction");
 812                 goto free_sring;
 813         }
 814
 815         return 0;
 816
 817 fail:
 818         xenbus_transaction_end(xbt, 1);
 819 free_sring:
 820         scsifront_free_ring(info);
 821
 822         return err;
 823 }
 824
 825
 826 static int scsifront_probe(struct xenbus_device *dev,
 827                            const struct xenbus_device_id *id)
 828 {
 829         struct vscsifrnt_info *info;
 830         struct Scsi_Host *host;
 831         int err = -ENOMEM;
 832         char name[TASK_COMM_LEN];
 833
 834         host = scsi_host_alloc(&scsifront_sht, sizeof(*info));
 835         if (!host) {
 836                 xenbus_dev_fatal(dev, err, "fail to allocate scsi host");
 837                 return err;
 838         }
 839         info = (struct vscsifrnt_info *)host->hostdata;
 840
 841         dev_set_drvdata(&dev->dev, info);
 842         info->dev = dev;
 843
 844         bitmap_fill(info->shadow_free_bitmap, VSCSIIF_MAX_REQS);
 845
 846         err = scsifront_init_ring(info);
 847         if (err) {
 848                 scsi_host_put(host);
 849                 return err;
 850         }
 851
 852         init_waitqueue_head(&info->wq_sync);
 853         init_waitqueue_head(&info->wq_pause);
 854         spin_lock_init(&info->shadow_lock);
 855
 856         snprintf(name, TASK_COMM_LEN, "vscsiif.%d", host->host_no);
 857
 858         host->max_id      = VSCSIIF_MAX_TARGET;
 859         host->max_channel = 0;
 860         host->max_lun     = VSCSIIF_MAX_LUN;
 861         host->max_sectors = (host->sg_tablesize - 1) * PAGE_SIZE / 512;
 862         host->max_cmd_len = VSCSIIF_MAX_COMMAND_SIZE;
 863
 864         err = scsi_add_host(host, &dev->dev);
 865         if (err) {
 866                 dev_err(&dev->dev, "fail to add scsi host %d\n", err);
 867                 goto free_sring;
 868         }
 869         info->host = host;
 870         info->host_active = 1;
 871
 872         xenbus_switch_state(dev, XenbusStateInitialised);
 873
 874         return 0;
 875
 876 free_sring:
 877         scsifront_free_ring(info);
 878         scsi_host_put(host);
 879         return err;
 880 }
 881
 882 static int scsifront_resume(struct xenbus_device *dev)
 883 {
 884         struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev);
 885         struct Scsi_Host *host = info->host;
 886         int err;
 887
 888         spin_lock_irq(host->host_lock);
 889
 890         /* Finish all still pending commands. */
 891         scsifront_finish_all(info);
 892
 893         spin_unlock_irq(host->host_lock);
 894
 895         /* Reconnect to dom0. */
 896         scsifront_free_ring(info);
 897         err = scsifront_init_ring(info);
 898         if (err) {
 899                 dev_err(&dev->dev, "fail to resume %d\n", err);
 900                 scsi_host_put(host);
 901                 return err;
 902         }
 903
 904         xenbus_switch_state(dev, XenbusStateInitialised);
 905
 906         return 0;
 907 }
 908
 909 static int scsifront_suspend(struct xenbus_device *dev)
 910 {
 911         struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev);
 912         struct Scsi_Host *host = info->host;
 913         int err = 0;
 914
 915         /* No new commands for the backend. */
 916         spin_lock_irq(host->host_lock);
 917         info->pause = 1;
 918         while (info->callers && !err) {
 919                 info->waiting_pause = 1;
 920                 info->wait_ring_available = 0;
 921                 spin_unlock_irq(host->host_lock);
 922                 wake_up(&info->wq_sync);
 923                 err = wait_event_interruptible(info->wq_pause,
 924                                                !info->waiting_pause);
 925                 spin_lock_irq(host->host_lock);
 926         }
 927         spin_unlock_irq(host->host_lock);
 928         return err;
 929 }
 930
 931 static int scsifront_remove(struct xenbus_device *dev)
 932 {
 933         struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev);
 934
 935         pr_debug("%s: %s removed\n", __func__, dev->nodename);
 936
 937         mutex_lock(&scsifront_mutex);
 938         if (info->host_active) {
 939                 /* Scsi_host not yet removed */
 940                 scsi_remove_host(info->host);
 941                 info->host_active = 0;
 942         }
 943         mutex_unlock(&scsifront_mutex);
 944
 945         scsifront_free_ring(info);
 946         scsi_host_put(info->host);
 947
 948         return 0;
 949 }
 950
 951 static void scsifront_disconnect(struct vscsifrnt_info *info)
 952 {
 953         struct xenbus_device *dev = info->dev;
 954         struct Scsi_Host *host = info->host;
 955
 956         pr_debug("%s: %s disconnect\n", __func__, dev->nodename);
 957
 958         /*
 959          * When this function is executed, all devices of
 960          * Frontend have been deleted.
 961          * Therefore, it need not block I/O before remove_host.
 962          */
 963
 964         mutex_lock(&scsifront_mutex);
 965         if (info->host_active) {
 966                 scsi_remove_host(host);
 967                 info->host_active = 0;
 968         }
 969         mutex_unlock(&scsifront_mutex);
 970
 971         xenbus_frontend_closed(dev);
 972 }
 973
 974 static void scsifront_do_lun_hotplug(struct vscsifrnt_info *info, int op)
 975 {
 976         struct xenbus_device *dev = info->dev;
 977         int i, err = 0;
 978         char str[64];
 979         char **dir;
 980         unsigned int dir_n = 0;
 981         unsigned int device_state;
 982         unsigned int hst, chn, tgt, lun;
 983         struct scsi_device *sdev;
 984
 985         dir = xenbus_directory(XBT_NIL, dev->otherend, "vscsi-devs", &dir_n);
 986         if (IS_ERR(dir))
 987                 return;
 988
 989         /* mark current task as the one allowed to modify device states */
 990         BUG_ON(info->curr);
 991         info->curr = current;
 992
 993         for (i = 0; i < dir_n; i++) {
 994                 /* read status */
 995                 snprintf(str, sizeof(str), "vscsi-devs/%s/state", dir[i]);
 996                 err = xenbus_scanf(XBT_NIL, dev->otherend, str, "%u",
 997                                    &device_state);
 998                 if (XENBUS_EXIST_ERR(err))
 999                         continue;
1000
1001                 /* virtual SCSI device */
1002                 snprintf(str, sizeof(str), "vscsi-devs/%s/v-dev", dir[i]);
1003                 err = xenbus_scanf(XBT_NIL, dev->otherend, str,
1004                                    "%u:%u:%u:%u", &hst, &chn, &tgt, &lun);
1005                 if (XENBUS_EXIST_ERR(err))
1006                         continue;
1007
1008                 /*
1009                  * Front device state path, used in slave_configure called
1010                  * on successfull scsi_add_device, and in slave_destroy called
1011                  * on remove of a device.
1012                  */
1013                 snprintf(info->dev_state_path, sizeof(info->dev_state_path),
1014                          "vscsi-devs/%s/state", dir[i]);
1015
1016                 switch (op) {
1017                 case VSCSIFRONT_OP_ADD_LUN:
1018                         if (device_state != XenbusStateInitialised)
1019                                 break;
1020
1021                         if (scsi_add_device(info->host, chn, tgt, lun)) {
1022                                 dev_err(&dev->dev, "scsi_add_device\n");
1023                                 err = xenbus_printf(XBT_NIL, dev->nodename,
1024                                               info->dev_state_path,
1025                                               "%d", XenbusStateClosed);
1026                                 if (err)
1027                                         xenbus_dev_error(dev, err,
1028                                                 "%s: writing dev_state_path", __func__);
1029                         }
1030                         break;
1031                 case VSCSIFRONT_OP_DEL_LUN:
1032                         if (device_state != XenbusStateClosing)
1033                                 break;
1034
1035                         sdev = scsi_device_lookup(info->host, chn, tgt, lun);
1036                         if (sdev) {
1037                                 scsi_remove_device(sdev);
1038                                 scsi_device_put(sdev);
1039                         }
1040                         break;
1041                 case VSCSIFRONT_OP_READD_LUN:
1042                         if (device_state == XenbusStateConnected) {
1043                                 err = xenbus_printf(XBT_NIL, dev->nodename,
1044                                               info->dev_state_path,
1045                                               "%d", XenbusStateConnected);
1046                                 if (err)
1047                                         xenbus_dev_error(dev, err,
1048                                                 "%s: writing dev_state_path", __func__);
1049                         }
1050                         break;
1051                 default:
1052                         break;
1053                 }
1054         }
1055
1056         info->curr = NULL;
1057
1058         kfree(dir);
1059 }
1060
1061 static void scsifront_read_backend_params(struct xenbus_device *dev,
1062                                           struct vscsifrnt_info *info)
1063 {
1064         unsigned int sg_grant, nr_segs;
1065         struct Scsi_Host *host = info->host;
1066
1067         sg_grant = xenbus_read_unsigned(dev->otherend, "feature-sg-grant", 0);
1068         nr_segs = min_t(unsigned int, sg_grant, SG_ALL);
1069         nr_segs = max_t(unsigned int, nr_segs, VSCSIIF_SG_TABLESIZE);
1070         nr_segs = min_t(unsigned int, nr_segs,
1071                         VSCSIIF_SG_TABLESIZE * PAGE_SIZE /
1072                         sizeof(struct scsiif_request_segment));
1073
1074         if (!info->pause && sg_grant)
1075                 dev_info(&dev->dev, "using up to %d SG entries\n", nr_segs);
1076         else if (info->pause && nr_segs < host->sg_tablesize)
1077                 dev_warn(&dev->dev,
1078                          "SG entries decreased from %d to %u - device may not work properly anymore\n",
1079                          host->sg_tablesize, nr_segs);
1080
1081         host->sg_tablesize = nr_segs;
1082         host->max_sectors = (nr_segs - 1) * PAGE_SIZE / 512;
1083 }
1084
1085 static void scsifront_backend_changed(struct xenbus_device *dev,
1086                                       enum xenbus_state backend_state)
1087 {
1088         struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev);
1089
1090         pr_debug("%s: %p %u %u\n", __func__, dev, dev->state, backend_state);
1091
1092         switch (backend_state) {
1093         case XenbusStateUnknown:
1094         case XenbusStateInitialising:
1095         case XenbusStateInitWait:
1096         case XenbusStateInitialised:
1097                 break;
1098
1099         case XenbusStateConnected:
1100                 scsifront_read_backend_params(dev, info);
1101
1102                 if (info->pause) {
1103                         scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_READD_LUN);
1104                         xenbus_switch_state(dev, XenbusStateConnected);
1105                         info->pause = 0;
1106                         return;
1107                 }
1108
1109                 if (xenbus_read_driver_state(dev->nodename) ==
1110                     XenbusStateInitialised)
1111                         scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_ADD_LUN);
1112
1113                 if (dev->state != XenbusStateConnected)
1114                         xenbus_switch_state(dev, XenbusStateConnected);
1115                 break;
1116
1117         case XenbusStateClosed:
1118                 if (dev->state == XenbusStateClosed)
1119                         break;
1120                 fallthrough;    /* Missed the backend's Closing state */
1121         case XenbusStateClosing:
1122                 scsifront_disconnect(info);
1123                 break;
1124
1125         case XenbusStateReconfiguring:
1126                 scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_DEL_LUN);
1127                 xenbus_switch_state(dev, XenbusStateReconfiguring);
1128                 break;
1129
1130         case XenbusStateReconfigured:
1131                 scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_ADD_LUN);
1132                 xenbus_switch_state(dev, XenbusStateConnected);
1133                 break;
1134         }
1135 }
1136
1137 static const struct xenbus_device_id scsifront_ids[] = {
1138         { "vscsi" },
1139         { "" }
1140 };
1141
1142 static struct xenbus_driver scsifront_driver = {
1143         .ids                    = scsifront_ids,
1144         .probe                  = scsifront_probe,
1145         .remove                 = scsifront_remove,
1146         .resume                 = scsifront_resume,
1147         .suspend                = scsifront_suspend,
1148         .otherend_changed       = scsifront_backend_changed,
1149 };
1150
1151 static int __init scsifront_init(void)
1152 {
1153         if (!xen_domain())
1154                 return -ENODEV;
1155
1156         return xenbus_register_frontend(&scsifront_driver);
1157 }
1158 module_init(scsifront_init);
1159
1160 static void __exit scsifront_exit(void)
1161 {
1162         xenbus_unregister_driver(&scsifront_driver);
1163 }
1164 module_exit(scsifront_exit);
1165
1166 MODULE_DESCRIPTION("Xen SCSI frontend driver");
1167 MODULE_LICENSE("GPL");
1168 MODULE_ALIAS("xen:vscsi");
1169 MODULE_AUTHOR("Juergen Gross <[email protected]>");